Cohort Analysis

What is Cohort Analysis?

Cohort analysis groups customers by the time period of their first purchase and then tracks their behavior over subsequent periods. Instead of looking at all customers as one pool, you see how each “generation” of buyers behaves as they age.

The key insight is that a single retention number (e.g., “30% of customers return”) hides whether things are getting better or worse. Cohort analysis reveals the trend: are newer cohorts retaining better than older ones, or are they churning faster?

Why It Matters

Aggregate metrics lie. If you’re acquiring more customers each month, a flat overall retention rate can mask the fact that newer cohorts are actually worse — the volume of new buyers is diluting the signal.

Cohort analysis answers the question that matters most for long-term business health: is the quality of your customer base improving over time? A business where Month 3 cohorts retain at 20% and Month 9 cohorts retain at 28% is on a fundamentally different trajectory than one where those numbers are reversed.

How to Read the Cohort Table

The table has a simple structure:

• Rows — each row is a cohort, identified by the month of first purchase (e.g., “Jan 2025”)
• Columns — each column is a period offset: Month 0 (first purchase month), Month 1 (one month later), Month 2, and so on
• Cells — the percentage of that cohort’s customers who made at least one purchase in that period

Read across a row to see how a single cohort ages. Read down a column to compare the same lifecycle stage across different cohorts. The diagonal represents the most recent data point for each cohort.

How AxonRow Builds Cohorts

AxonRow identifies each buyer by their buyer_uid from TikTok Shop order data. The first order date for each buyer determines their cohort assignment. Subsequent orders from the same buyer are counted as retention events in the appropriate month offset.

Orders are only counted once per buyer per period — a buyer who places three orders in Month 2 still counts as one retained customer for that cell. This gives you a clean picture of how many customers are still active, not how many orders they’re placing.

What Good Looks Like

Retention benchmarks vary by product category, but for TikTok Shop sellers:

• Month 1 retention of 15–25% is typical for consumable or replenishable products (skincare, supplements, household goods)
• Month 1 retention below 10% is common for one-time purchases (electronics accessories, novelty items) and is not necessarily a problem if LTV is priced in
• Improving cohort curves — where newer cohorts show higher retention at the same month offset — indicate that product quality, post-purchase experience, or targeting is improving
• Flat or declining curves — where newer cohorts perform worse — are an early warning sign worth investigating before the trend compounds

A healthy business typically sees Month 1 retention stabilize or improve over time, with Month 3+ retention forming a floor that represents your loyal repeat-buyer base.

Common Mistakes

• Comparing cohorts of different sizes without normalizing — A 500-buyer cohort and a 50-buyer cohort both show percentages, but the smaller cohort’s numbers are noisier. Look for trends across multiple cohorts rather than reacting to a single data point.
• Confusing cohort retention with overall retention rate — The dashboard’s overall retention metric blends all cohorts together. Use the cohort table when you want to understand whether retention is structurally improving, not just what it is today.
• Acting on incomplete cohorts — The most recent 1–2 cohorts don’t have enough history to draw conclusions. Wait until a cohort is at least 3 months old before using it to inform product or marketing decisions.